I’ve seen this movie before. That’s why a recent LinkedIn post by Ilya Kabanov stopped me mid-doomscroll.

Kabanov described how frontier AI companies are quietly but decisively shifting into cybersecurity. They are not joining as partners or tacking on features. They are stepping up as product makers, targeting the core of the enterprise security budget. This was not some crafted press release or hype. It was sharp pattern recognition.

If you have been in tech long enough, this pattern rings familiar. Years ago, when NoSQL databases took off in the enterprise, I had a candid conversation with the CEOs of MongoDB and Couchbase. I asked if NoSQL meant “no security.” Their answer was honest: Security was not the meaningful obstacle. Customers were. The message was clear. Security would come when customers demanded it. Until then, velocity, adoption, and developer excitement ruled.

This is how new waves in tech play out. Adoption comes first. Security arrives when buyers finally get loud about it.

This is exactly what we are seeing right now in AI.

The Shift: AI’s “Customers Want Security” Moment

Early AI models sprinted to market at incredible speed. Vendors focused on getting features out the door. Safety measures were put off for another day. Today, the conversation has changed. Enterprises are applying real pressure. Boards want to see risk management plans, not just AI output. CISOs are asking tougher questions. The market expects security built into the DNA of AI itself.

Recent survey work from the Futurum Group reinforces this shift. In the Security & AI Decision Maker Pulse for Q1 2026, improving security and transparency for AI models was the top procurement concern for enterprise buyers. More than four out of five CISOs now say protecting AI models and data is just as important as classic perimeter security. Two years back, that was less than half.

This Isn’t Just Testing the Waters

Kabanov’s breakdown of the field says it all. OpenAI, Anthropic, and Google DeepMind are building core security features as mainline products. OpenAI has kicked off a security business unit to turn internal tools into offerings for customers. Anthropic has put a former SentinelOne leader at the head of its AI security team. DeepMind is developing CodeMender for autonomous vulnerability fixes, and plugging new tech into Google’s mature security stack. These are not pilot projects. They are major bets.

Futurum’s Global Cybersecurity Landscape Report for 2026 projects AI-powered, autonomous security solutions will account for about $96 billion in enterprise investment next year. That comes out to more than a third of all spending on cybersecurity platforms. The same study found a 40% jump in RFPs requiring “AI-native, embedded security” in the past twelve months. The change is being driven by what buyers now demand, not a mere vendor push.

xAI is quiet, for now. But with the global market for security and resilience heading for $260 billion by the end of 2026, that silence won’t last.

“Secure by Default” Finally Means Something

For years, “secure by default” was mostly empty talk. People would buy the primary platform, add security after-the-fact, then add tools to manage that pileup. Complexity increased, but risk often stayed put. The rise of autonomous, AI-driven security flips this formula. Security is becoming a property of the system itself. Not a checkbox to bolt on, but a standard outcome for the buyer.

Futurum’s 2026 State of Platform Security study found that 68% of large enterprises are working to shrink and simplify their security stack. The average number of separate security products in use is down by 40% since 2024. Buyers now see simple detection as inadequate. They want security that automatically detects and fixes, not just alerts.

The Business Model for Security Is Shifting

Embedded security changes how protection is delivered and how it is paid for. Standalone tool makers now need a stronger case to survive. Buyers expect robust, secure defaults. The shift to AI-native security means vendors who leaned on selling security as an add-on could see their margins squeezed in the years ahead.

This is not the first time the tech industry has gone from “we’ll secure it when customers care” to “they care right now, and we need to deliver.” But it may be the fastest. The next year or two will show who adapts and who disappears. AI’s velocity is now being matched by hard questions about safety and trust. This time, customers are not just asking for security. They are demanding it—and the smartest players already know the answer is yes.